wav2vec2-xls-r-300m-nyn_filtered-yogera-v1

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.7382
• Model Preparation Time: 0.0103
• Wer: 0.5477
• Cer: 0.1679

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0003
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.033
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Wer	Cer
7.5062	1.0	137	2.9944	0.0103	1.0	1.0
2.8439	2.0	274	2.7840	0.0103	1.0	1.0
1.2957	3.0	411	0.9259	0.0103	0.7567	0.2434
0.4792	4.0	548	0.8136	0.0103	0.6549	0.2037
0.3563	5.0	685	0.8125	0.0103	0.6330	0.1982
0.2934	6.0	822	0.7634	0.0103	0.6229	0.2013
0.2445	7.0	959	0.7673	0.0103	0.6129	0.1899
0.2069	8.0	1096	0.7700	0.0103	0.6323	0.1977
0.1773	9.0	1233	0.8577	0.0103	0.6233	0.1916
0.1539	10.0	1370	0.8454	0.0103	0.6169	0.1916
0.1252	11.0	1507	0.9611	0.0103	0.6181	0.1904
0.1078	12.0	1644	1.0559	0.0103	0.6321	0.1911
0.0938	13.0	1781	1.0658	0.0103	0.6244	0.1928
0.0806	14.0	1918	1.0884	0.0103	0.6163	0.1879
0.0713	15.0	2055	1.0535	0.0103	0.6162	0.1864
0.0638	16.0	2192	1.0985	0.0103	0.6159	0.1893
0.0608	17.0	2329	1.1801	0.0103	0.6078	0.1887
0.0528	18.0	2466	1.1715	0.0103	0.6070	0.1861
0.0483	19.0	2603	1.2085	0.0103	0.6221	0.1865
0.0441	20.0	2740	1.2510	0.0103	0.6057	0.1842
0.0416	21.0	2877	1.2759	0.0103	0.6058	0.1845
0.0432	22.0	3014	1.3016	0.0103	0.5996	0.1840
0.0371	23.0	3151	1.2451	0.0103	0.5957	0.1831
0.036	24.0	3288	1.3475	0.0103	0.6098	0.1889
0.0373	25.0	3425	1.3265	0.0103	0.6069	0.1893
0.0346	26.0	3562	1.3153	0.0103	0.5912	0.1826
0.0335	27.0	3699	1.3640	0.0103	0.5987	0.1896
0.0322	28.0	3836	1.2965	0.0103	0.5902	0.1830
0.0286	29.0	3973	1.4187	0.0103	0.5951	0.1844
0.0284	30.0	4110	1.3861	0.0103	0.5864	0.1810
0.0274	31.0	4247	1.3071	0.0103	0.5843	0.1782
0.0291	32.0	4384	1.3692	0.0103	0.5829	0.1810
0.0273	33.0	4521	1.3644	0.0103	0.5839	0.1820
0.0271	34.0	4658	1.3298	0.0103	0.5909	0.1817
0.0252	35.0	4795	1.3897	0.0103	0.5787	0.1794
0.0257	36.0	4932	1.4339	0.0103	0.5891	0.1823
0.0236	37.0	5069	1.4054	0.0103	0.5976	0.1845
0.0239	38.0	5206	1.3777	0.0103	0.5831	0.1810
0.0227	39.0	5343	1.3901	0.0103	0.5857	0.1811
0.0233	40.0	5480	1.3737	0.0103	0.5904	0.1833
0.0234	41.0	5617	1.3843	0.0103	0.5843	0.1821
0.0217	42.0	5754	1.3170	0.0103	0.5822	0.1811
0.0197	43.0	5891	1.4165	0.0103	0.5847	0.1800
0.0197	44.0	6028	1.4449	0.0103	0.5759	0.1773
0.0203	45.0	6165	1.3591	0.0103	0.5843	0.1790
0.0199	46.0	6302	1.5098	0.0103	0.5840	0.1806
0.0196	47.0	6439	1.4038	0.0103	0.5755	0.1767
0.0198	48.0	6576	1.4440	0.0103	0.5798	0.1779
0.0194	49.0	6713	1.4583	0.0103	0.5837	0.1782
0.0179	50.0	6850	1.4186	0.0103	0.5738	0.1774
0.0166	51.0	6987	1.4620	0.0103	0.5735	0.1777
0.0163	52.0	7124	1.5091	0.0103	0.5691	0.1761
0.0167	53.0	7261	1.4249	0.0103	0.5723	0.1759
0.0156	54.0	7398	1.5054	0.0103	0.5773	0.1791
0.0165	55.0	7535	1.4583	0.0103	0.5812	0.1796
0.0136	56.0	7672	1.5845	0.0103	0.5710	0.1766
0.015	57.0	7809	1.4338	0.0103	0.5750	0.1769
0.0139	58.0	7946	1.6348	0.0103	0.5789	0.1790
0.0141	59.0	8083	1.5682	0.0103	0.5784	0.1773
0.0135	60.0	8220	1.5523	0.0103	0.5678	0.1741
0.0138	61.0	8357	1.5624	0.0103	0.5730	0.1768
0.0125	62.0	8494	1.5838	0.0103	0.5707	0.1749
0.0126	63.0	8631	1.5254	0.0103	0.5614	0.1742
0.0116	64.0	8768	1.6224	0.0103	0.5606	0.1740
0.0127	65.0	8905	1.5903	0.0103	0.5678	0.1761
0.0115	66.0	9042	1.5875	0.0103	0.5672	0.1747
0.0116	67.0	9179	1.6402	0.0103	0.5684	0.1761
0.0114	68.0	9316	1.6358	0.0103	0.5666	0.1756
0.0111	69.0	9453	1.5798	0.0103	0.5607	0.1738
0.011	70.0	9590	1.6475	0.0103	0.5714	0.1771
0.0119	71.0	9727	1.5381	0.0103	0.5704	0.1779
0.0112	72.0	9864	1.5897	0.0103	0.5646	0.1741
0.0105	73.0	10001	1.6031	0.0103	0.5614	0.1721
0.0106	74.0	10138	1.5518	0.0103	0.5708	0.1747
0.0102	75.0	10275	1.5620	0.0103	0.5637	0.1748
0.0092	76.0	10412	1.6083	0.0103	0.5644	0.1750
0.0098	77.0	10549	1.6316	0.0103	0.5626	0.1735
0.0084	78.0	10686	1.6316	0.0103	0.5537	0.1709
0.0086	79.0	10823	1.6213	0.0103	0.5577	0.1728
0.008	80.0	10960	1.6382	0.0103	0.5518	0.1720
0.008	81.0	11097	1.6010	0.0103	0.5515	0.1702
0.0079	82.0	11234	1.6379	0.0103	0.5549	0.1716
0.0078	83.0	11371	1.6614	0.0103	0.5559	0.1716
0.0078	84.0	11508	1.6183	0.0103	0.5611	0.1732
0.0076	85.0	11645	1.6835	0.0103	0.5524	0.1707
0.0064	86.0	11782	1.6764	0.0103	0.5515	0.1697
0.0067	87.0	11919	1.6797	0.0103	0.5548	0.1694
0.0065	88.0	12056	1.6550	0.0103	0.5503	0.1697
0.0064	89.0	12193	1.7282	0.0103	0.5504	0.1695
0.0062	90.0	12330	1.6935	0.0103	0.5506	0.1701
0.0061	91.0	12467	1.7180	0.0103	0.5576	0.1704
0.0057	92.0	12604	1.7489	0.0103	0.5521	0.1689
0.0057	93.0	12741	1.7571	0.0103	0.5537	0.1681
0.0058	94.0	12878	1.7497	0.0103	0.5517	0.1682
0.0057	95.0	13015	1.7382	0.0103	0.5477	0.1679
0.0054	96.0	13152	1.7279	0.0103	0.5491	0.1675
0.0055	97.0	13289	1.7383	0.0103	0.5500	0.1679
0.0052	98.0	13426	1.7410	0.0103	0.5484	0.1676
0.0054	99.0	13563	1.7224	0.0103	0.5484	0.1671
0.0051	100.0	13700	1.7312	0.0103	0.5507	0.1674

Framework versions

• Transformers 4.44.0
• Pytorch 2.1.0+cu118
• Datasets 2.20.0
• Tokenizers 0.19.1